Komodo 12.1

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
S.Taylor
Posts: 7824
Joined: Thu Mar 09, 2006 2:25 am
Location: Jerusalem Israel

Re: Komodo 12.1

Post by S.Taylor » Wed Jun 06, 2018 9:56 pm

Komodo is doing very well in TCEC at this momment. Which ever version, it would be a shame not to see if it can win this season of tcec.
Next season we can have it win crushingly.

S.Taylor
Posts: 7824
Joined: Thu Mar 09, 2006 2:25 am
Location: Jerusalem Israel

Re: Komodo 12.1

Post by S.Taylor » Mon Jun 18, 2018 7:46 pm

Last post was 6th of june.
Since then, komodo achieved a huge victory imho. Tcec.
Tomorrow i think, the tcec superfinal begins.
Surely in light of this, the komodo team would have been working feverishly to add a few more elo to the main engine?
Has this been the case?

User avatar
CMCanavessi
Posts: 599
Joined: Thu Dec 28, 2017 3:06 pm
Location: Argentina

Re: Komodo 12.1

Post by CMCanavessi » Mon Jun 18, 2018 8:01 pm

TCEC final has already started
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls

mjlef
Posts: 1330
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: Komodo 12.1

Post by mjlef » Mon Jun 18, 2018 8:54 pm

S.Taylor wrote:
Mon Jun 18, 2018 7:46 pm
Last post was 6th of june.
Since then, komodo achieved a huge victory imho. Tcec.
Tomorrow i think, the tcec superfinal begins.
Surely in light of this, the komodo team would have been working feverishly to add a few more elo to the main engine?
Has this been the case?
Most of the Komodo 12.1.1 update were improvements to MCTS mode, which is not being used in TCEC. But there was on small extension related change worth maybe 3-4- elo. Larry also found a higher Dynamism improved ply about 20 elo on 12 cores, so we are running with that, With more testing will work on an automatic Dynamism based on number of cores for the next Komodo release.

S.Taylor
Posts: 7824
Joined: Thu Mar 09, 2006 2:25 am
Location: Jerusalem Israel

Re: Komodo 12.1

Post by S.Taylor » Mon Jun 18, 2018 9:04 pm

CMCanavessi wrote:
Mon Jun 18, 2018 8:01 pm
TCEC final has already started
Wow it sure HAS!
Game 1 underway, not normal!

Dann Corbit
Posts: 8674
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Komodo 12.1

Post by Dann Corbit » Mon Jun 18, 2018 9:51 pm

K won first game.
Second is started.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

Leo
Posts: 459
Joined: Fri Sep 16, 2016 4:55 pm
Location: Saint Paul, Minnesota USA

Re: Komodo 12.1

Post by Leo » Mon Jun 18, 2018 9:54 pm

Quote from the main page of the TCEC superfinal: " Development is still in progress. Not everything is 100% complete. Whiners will be ejected."
Brainfish fan. AMD fan.

User avatar
Nordlandia
Posts: 2037
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Komodo 12.1

Post by Nordlandia » Tue Jun 19, 2018 6:16 am

I like the new live interface. Old website has to be alteast 5-6 years by now.

User avatar
Nordlandia
Posts: 2037
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Komodo 12.1

Post by Nordlandia » Tue Jun 19, 2018 3:21 pm

In case of K12.1.1 selfplay, which contempt value is recommended?

Last time i asked it was at 13. It may have chanced since then, thereby asking.

I kindly request recommended values for Komodo 12.1.1 vs Komodo 12.1.1 and Komodo 12.1.1 versus Komodo 11.3.1

#1 K12.1.1 identical selfplay

#2 K12.1.1 face previous official release (Komodo 11.3.1)

lkaufman
Posts: 3120
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 12.1

Post by lkaufman » Wed Jun 20, 2018 5:12 am

Nordlandia wrote:
Tue Jun 19, 2018 3:21 pm
In case of K12.1.1 selfplay, which contempt value is recommended?

Last time i asked it was at 13. It may have chanced since then, thereby asking.

I kindly request recommended values for Komodo 12.1.1 vs Komodo 12.1.1 and Komodo 12.1.1 versus Komodo 11.3.1

#1 K12.1.1 identical selfplay

#2 K12.1.1 face previous official release (Komodo 11.3.1)
For selfplay, if one side uses contempt and the other uses zero, the zero version should come out ahead, if only by a couple elo. But there is an advantage to using the normal contempt value for both sides of selfplay, namely that you get less draws. So when we test a minor possible improvement vs. the version without it, we use normal Contempt to reduce draws. For 12.1.1 vs 11.3.1 theoretically the best result would be with a tiny positive Contempt like 1 or 2, but the effect is too tiny to measure. The next version will probably allow Contempt to be used for the MCTS version; we used Contempt 50 vs Nakamura.
Komodo rules!

Post Reply